《系统工程》课程教学资源（英文文献）Hierarchical modeling and simulation of dynamic production logistic system.pdf_大学文库

Hierarchical modeling and simulation of dynamicproduction logistic systemGuide Words: production logistic system; hierarchical model; dynamic modelerAbstract: Aimed to the problem of model bound with data and structure rigidity in simulation,hierarchical modelling and simulation technology for dynamical production logistic system ispresented with the view of data-driven and dynamical hierarchy relating. Representative model of fourcontrol levels including factory(FM), workshop(SM), cell(CM) and equipment(EM) and its dynamicmodeler(DM)wasputforward.Data-drivenmodelgeneration andrunningmechanismwas explainedto solve the automatic modeling problem of fast application and system integration.Flexiblesimulation control mechanismbased on dynamic layerrelatingwas constructed by usingBUFFERand AGV. Realization rules and conditions of dynamic layer relating was analyzed in detail, and itsprocedure parsed by examples.Development and implementation indicates the practical values of thehierarchical modeling and simulation technology of the dynamic production logistic system.L.INTRODUCTIONRecently, material logistic modernization has been gathering more and more concern byenterprisers, for they have to encounter increasing pressure from market under increasing internationalcompetition. In enterprise, logistic activities for production are called production logistics. Productionlogistic system (PLS) is a complex system characterized by discretization and random city, in whichlots of optimization problems have to be solved in system design and control and it's difficult toobtain optimum solution or satisfactory solution using traditional analytic methods. As an effectiveenabling technology in system analyses and research, Modeling & Simulation (M&S) has been widelyapplied into PLS in planning design, transportation schedule and material control, etc. In simulation,modeling is thekey.Model is thefoundation of simulation,and simulation is the real run of model oncomputer. For its compatible characters to object orientation, modularization and flexibilityHierarchical modeling have gained its position in numerous modeling technologies and manyconstitutions and scholars have been putting their effort in this field. America National StandardsCommittee (ANSC) classifies manufacturing system into five levels as enterprise, factory, cellworkstation and machine. In IsO, enterprise automatization system is composed of factory, workshop.cell, workstation and machine. Cellular manufacturing system modeled as hierarchical levels ofworkshop, cell and machine is studied by ZHANG Lunyan ; XUE Jiabing disparts FMS simulationcontrol structure into FMS, cell and machine; Venkates waran Jpresents a hybrid simulation

Hierarchical modeling and simulation of dynamic production logistic system Guide Words：production logistic system; hierarchical model; dynamic modeler Abstract：Aimed to the problem of model bound with data and structure rigidity in simulation, hierarchical modelling and simulation technology for dynamical production logistic system is presented with the view of data-driven and dynamical hierarchy relating. Representative model of four control levels including factory(FM), workshop(SM), cell(CM) and equipment(EM) and its dynamic modeler (DM) was put forward. Data-driven model generation and running mechanism was explained to solve the automatic modeling problem of fast application and system integration. Flexible simulation control mechanism based on dynamic layer relating was constructed by using BUFFER and AGV. Realization rules and conditions of dynamic layer relating was analyzed in detail, and its procedure parsed by examples. Development and implementation indicates the practical values of the hierarchical modeling and simulation technology of the dynamic production logistic system. I. INTRODUCTION Recently, material logistic modernization has been gathering more and more concern by enterprisers, for they have to encounter increasing pressure from market under increasing international competition. In enterprise, logistic activities for production are called production logistics. Production logistic system (PLS) is a complex system characterized by discretization and random city, in which lots of optimization problems have to be solved in system design and control and it’s difficult to obtain optimum solution or satisfactory solution using traditional analytic methods. As an effective enabling technology in system analyses and research, Modeling & Simulation (M&S) has been widely applied into PLS in planning design, transportation schedule and material control, etc. In simulation, modeling is the key. Model is the foundation of simulation, and simulation is the real run of model on computer. For its compatible characters to object orientation, modularization and flexibility, Hierarchical modeling have gained its position in numerous modeling technologies and many constitutions and scholars have been putting their effort in this field. America National Standards Committee (ANSC) classifies manufacturing system into five levels as enterprise, factory, cell, workstation and machine. In ISO, enterprise automatization system is composed of factory, workshop, cell, workstation and machine. Cellular manufacturing system modeled as hierarchical levels of workshop, cell and machine is studied by ZHANG Lunyan ; XUE Jiabing disparts FMS simulation control structure into FMS, cell and machine; Venkates waran Jpresents a hybrid simulation

production planning system based on enterprise and workshop simulators, in which every layer hasfeedbackcontrol ringtomonitor systemperformanceandupdatecontrolparameters;LarsMonchdescribes the match of multi-agent model into sophisticated workshop production control that is madeup of enterprise, workgroup and machine. Referring to hierarchical structures mentioned above, fromthe viewpoint of object orientation and modularization, a rapid hierarchical PLS model suitable tomajority enterprise is constructed here.Then creation method of dynamic hierarchical model isstudied usingthethoughtsofflexible and dynamicmodeling.Theultimatemotive isto quickensystem modeling, improve efficiency and effect of logistic simulation,and enlarge applicable scope ofhierarchical model.II.PROBLEMANALYSESHierarchical model is abstractedfrom real department and business structure and havehightechnology adaptability that it can well satisfy modeling requirements of object orientation, dynamicflexibility, intelligence, etc. so it is widely used in various modeling technologies. For this reason,hierarchical model can be understood as a common modeling thoughts or model form. It is a real pitythat applicable scope of hierarchical model is so limited and its applicable effect is not remarkableamong existing research results.Reasons behind are summarized here:(1) Model and data are confused. Such applications create system model according to enterpriseor workshop hierarchical structure, but its scope is limited to specific project, in which systemgeneralization is not emphasized and model is bound together with its related data during modelconstruction.These instances will inevitably produce problems of indistinct hierarchy and data flowconfusion.Professional team have to be settled in during implementation and full-time maintenancehave to be paid when system is running. Finally long implementation cycle and high maintenance costhavetobepaid.Asanprofessional applicationtechnologyinPLS,M&Sshouldnotbeabottleneck insystem implementation and running.(2) Model structure is rigid. Same model cannot adapt to different production environment.According to the theory of business process reconfiguration, complexity of multi-layer hierarchicalcontrol system is the cause of flexibility lacking in workshop and response lagging to internal andexternal changes. complicated and rigid model structure cannot fit rapid modeling requirement, thusapplication to other enterprises is not realistic. Traditional models are static and their layers of arechangeless.In practice, requirements for hierarchical structure will vary in enterprises or departments in oneenterprise and even application phases of same department.Specific phenomenon as scopescale-down, skipping and hybrid layer exists widely. Typical example is that machines are laid directlyunder factory, where no concept of cell is applied. Another example is that some machining center canbe regarded as work cell in workshop.Then what kind of model form or modeling method cansuitably exhibit its generalization of hierarchical model and how about its rapid and dynamic forming

production planning system based on enterprise and workshop simulators, in which every layer has feedback control ring to monitor system performance and update control parameters; Lars Monch describes the match of multi-agent model into sophisticated workshop production control that is made up of enterprise, workgroup and machine. Referring to hierarchical structures mentioned above, from the viewpoint of object orientation and modularization, a rapid hierarchical PLS model suitable to majority enterprise is constructed here. Then creation method of dynamic hierarchical model is studied using the thoughts of flexible and dynamic modeling. The ultimate motive is to quicken system modeling, improve efficiency and effect of logistic simulation, and enlarge applicable scope of hierarchical model. II. PROBLEM ANALYSES Hierarchical model is abstracted from real department and business structure and have high technology adaptability that it can well satisfy modeling requirements of object orientation, dynamic flexibility, intelligence, etc. so it is widely used in various modeling technologies. For this reason, hierarchical model can be understood as a common modeling thoughts or model form. It is a real pity that applicable scope of hierarchical model is so limited and its applicable effect is not remarkable among existing research results. Reasons behind are summarized here: (1) Model and data are confused. Such applications create system model according to enterprise or workshop hierarchical structure, but its scope is limited to specific project, in which system generalization is not emphasized and model is bound together with its related data during model construction. These instances will inevitably produce problems of indistinct hierarchy and data flow confusion. Professional team have to be settled in during implementation and full-time maintenance have to be paid when system is running. Finally long implementation cycle and high maintenance cost have to be paid. As an professional application technology in PLS, M&S should not be a bottleneck in system implementation and running. (2) Model structure is rigid. Same model cannot adapt to different production environment. According to the theory of business process reconfiguration, complexity of multi-layer hierarchical control system is the cause of flexibility lacking in workshop and response lagging to internal and external changes. complicated and rigid model structure cannot fit rapid modeling requirement, thus application to other enterprises is not realistic. Traditional models are static and their layers of are changeless. In practice, requirements for hierarchical structure will vary in enterprises or departments in one enterprise and even application phases of same department. Specific phenomenon as scope scale-down, skipping and hybrid layer exists widely. Typical example is that machines are laid directly under factory, where no concept of cell is applied. Another example is that some machining center can be regarded as work cell in workshop. Then what kind of model form or modeling method can suitably exhibit its generalization of hierarchical model and how about its rapid and dynamic forming

character. Few efforts are involved, and existing studies are not comprehensive and profound. So thispaperwill startwithaspectsasfollows.(1) Aimed to rapid construction, data-driven static hierarchical model is to be establishedaccording to the thoughts of separation of model and data. Thus, build-time and run-time view isintroduced to take model construction part with simulation run. During build-time, basic user objectsare enveloped and developed, and their relations are automatically created under driving of externaldata, which can also drive automatic execution of simulation during runtime.(2) Dynamic hierarchical model will be researched to satisfy the requirement of dynamic andflexible modeling. Phenomenon of scope scale-down, skipping and hybrid layer in hierarchical modelcan be concluded as the problem of layer relating. Based on static model, layer relating objects insimulation software can befully used to solvethis problem,In traditional PLShierarchical modelmodel layers are related statically, work pieces transported from source to destination machine areuploaded and downloaded layer by layer according to fixed path, During uploading, work pieces areloaded from bufferof source layer, whilethey have tobe uninstalled into buffer ofdestination layerduringdownloading.Buffershere act as an important role in static layerrelating.But in dynamichierarchical model,model layers havenotfixed relationthat theyhavethetrend to beflatted withenterprise department, so they can only be viewed as abstract layers like up, middle and under layer.Relation of abstract layers can be easily understood and realized. Dynamic factors also result indynamically judgment of work piece transportation path, AGV (Auto Guided Vehicle) equipment'sand buffers together are appropriate objects to bear the role of dynamic relating. Concentrating onabove problems and technical difficulties, this paper will take PLS as the object of study, andimplement dynamic hierarchical model recurring to powerful capacities of modeling and secondarydevelopmentofcommercialsimulationplatform.IIL.OVERALLTECHNICALSCHEMEThrough problem analyses, the purpose of this paper is affirmed, that is rapid construction ofhierarchical model and its dynamic flexible realization. Therefore, generation and running mechanismof data-driven model and its dynamic realization is selected as emphases of this paper. Fig. 1illustrates the detail technical scheme

character. Few efforts are involved, and existing studies are not comprehensive and profound. So this paper will start with aspects as follows. (1) Aimed to rapid construction, data-driven static hierarchical model is to be established according to the thoughts of separation of model and data. Thus, build-time and run-time view is introduced to take model construction part with simulation run. During build-time, basic user objects are enveloped and developed, and their relations are automatically created under driving of external data, which can also drive automatic execution of simulation during runtime. (2) Dynamic hierarchical model will be researched to satisfy the requirement of dynamic and flexible modeling. Phenomenon of scope scale-down, skipping and hybrid layer in hierarchical model can be concluded as the problem of layer relating. Based on static model, layer relating objects in simulation software can be fully used to solve this problem. In traditional PLS hierarchical model, model layers are related statically, work pieces transported from source to destination machine are uploaded and downloaded layer by layer according to fixed path. During uploading, work pieces are loaded from buffer of source layer, while they have to be uninstalled into buffer of destination layer during downloading. Buffers here act as an important role in static layer relating. But in dynamic hierarchical model, model layers have not fixed relation that they have the trend to be flatted with enterprise department, so they can only be viewed as abstract layers like up, middle and under layer. Relation of abstract layers can be easily understood and realized. Dynamic factors also result in dynamically judgment of work piece transportation path, AGV (Auto Guided Vehicle) equipment’s and buffers together are appropriate objects to bear the role of dynamic relating. Concentrating on above problems and technical difficulties, this paper will take PLS as the object of study, and implement dynamic hierarchical model recurring to powerful capacities of modeling and secondary development of commercial simulation platform. III. OVERALL TECHNICAL SCHEME Through problem analyses, the purpose of this paper is affirmed, that is rapid construction of hierarchical model and its dynamic flexible realization. Therefore, generation and running mechanism of data-driven model and its dynamic realization is selected as emphases of this paper. Fig. 1 illustrates the detail technical scheme

Dynamic layer realization based onModel generation and running byBUFFERs andAGVdata-drivenExtemaldataDynamicBuffermodelingDataimportingmyorData interfinceInternal datappingarnBufferMiddleridlaverData-drivenDlayesBasic modelingobjectsDbjecDynamicUsermodelingBuffersimulationobjectsUndelaverRun-timeview-Build-time vieFig.1 overall technical schemeFig.1 divides model formation and running respectively into build-time and run-time view. Modelformation isrealized mainlythroughdevelopment of Dynamic Modeller(DM)usingdata-drivenmodelingmethod.InDM,externaldataisconvertedintoformalizedinternaldatabydatainterface:basicobjectsareenveloped intousermodelingobjects,whichwill berelated into usermodel bydata-driven mechanism. Data-driven user model is really a kind of variable metric adjustable modelthat is decided by related data input. In fig.l, if the middle layer is drawn out, then the up layer willdescent to fill up its position so that the new model appears to be layer-skipped. In this way modelmetric gets adjusted. In simulation software, buffers can be used as relating entity, model layer logic issealed in it and can be parsed through data-driven method using information from layout, tasks andprocesses. So, from the point of data flow, buffers can also be seemed as data mapping zones.Model created with layer relating information sealed in buffers in build-time view is static.Dynamic layer relating have to be implemented in run-time view through transportation equipmentlike AGV. During the procedure of loading, transporting and unloading, AGV trailer will use layerinformation inbuffers and sourceand destination machine of workpiecetodecideits routeIV.KEYTECHNOLOGIESAccording to management and operation model of most enterprise, four layer structure ispresented (Table 1). User objects enveloped includes FM, SM, CM, EM, BUFFER and AGV. Theirlayer relation are also illustrated in table 1, in which the object of BUFFER is sealed in different mainlayers to household work pieces and relate layers, and AGV exists in layers upper than EM to finishtransportation actions and realize dynamic layer relating. In PLS, presence of AGV in simulation willsignificantly enhancethe effectofmodel running to someextent.Table 1: layer objects and their relationship

Fig.1 overall technical scheme Fig.1 divides model formation and running respectively into build-time and run-time view. Model formation is realized mainly through development of Dynamic Modeller (DM) using data-driven modeling method. In DM, external data is converted into formalized internal data by data interface; basic objects are enveloped into user modeling objects, which will be related into user model by data-driven mechanism. Data-driven user model is really a kind of variable metric adjustable model that is decided by related data input. In fig.1, if the middle layer is drawn out, then the up layer will descent to fill up its position so that the new model appears to be layer-skipped. In this way model metric gets adjusted. In simulation software, buffers can be used as relating entity, model layer logic is sealed in it and can be parsed through data-driven method using information from layout, tasks and processes. So, from the point of data flow, buffers can also be seemed as data mapping zones. Model created with layer relating information sealed in buffers in build-time view is static. Dynamic layer relating have to be implemented in run-time view through transportation equipment like AGV. During the procedure of loading, transporting and unloading, AGV trailer will use layer information in buffers and source and destination machine of work piece to decide its route. IV. KEY TECHNOLOGIES According to management and operation model of most enterprise, four layer structure is presented (Table 1). User objects enveloped includes FM, SM, CM, EM, BUFFER and AGV. Their layer relation are also illustrated in table 1, in which the object of BUFFER is sealed in different main layers to household work pieces and relate layers, and AGV exists in layers upper than EM to finish transportation actions and realize dynamic layer relating. In PLS, presence of AGV in simulation will significantly enhance the effect of model running to some extent. Table 1: layer objects and their relationship

tasks,productprocesses andresourcesintheformerinstanceareorganizedinternallyorimportedfrom independentdatabase, while in the latter data havetobe imported from related system or bymiddle database. Like other integration applications, data mapping technologies are employed totransform external data into internal format, such translating work is done in DM. Main data tableused inPLSarelistedinTable2.Readerscanrefertofig.4toseetheirformat and instancesTable 2: formatted internal data tableTablenameDataData formattingsourceMODINFOFM/SM/CM/EM,objectcooLayout,resourcesrdinate, resources statusTRACKINFOLayout,AGV,object coordinate, resresourcesourcesstatusMODLINKLayer info. for objects in MLayoutODINFOTaskListProductionBatch, planned and actual prtasksoduction/deliverytimeMasterTableNested relation table includiProductprocessesngroute,process,optionalmachine with machining time.Information saved in MODINFO, TRACKINFO and MODLINK is used for generatinghierarchical structureof usermodel in build-time view.Coordinate and status of resources,tasks andprocesses are dedicated to model running in run-time view.Themostearliestproductiontimeandthemost latest deliverytime intaskscanbe seemed astheactual start and end running time of simulation respectively. Work pieces in tasks are created in toplayer, then they are transferred directly layer by layer along selected routes decided by control rules,or transported by AGV if it is considered in user model; work pieces arrive at machining equipment(EM) and are stored in BUFFERs waiting for machining, The machine selects work piece formachining via selection rules, After machining is finished, the part will be uploaded layer by layer andin the same way downloaded to destination machine responding for its next process.Generally, division of labor in workshop in factory or in cell in workshop is clear, and assignmentof task and material transportation route is decided. But when task arrived at cell layer, selection ofmachining equipment is uncertain according to different schedule rules. Thus, in static model, materialtransportation is mainly realized in up layers (FM and SM), while middle layer (CM) is used for taskassignment, and under layer (EM) only responds for machining.This character can also be adapted todynamic model. Hence, logic of task assignment of material transportation and machining can beembedded in the joint of layers, that can be easily realized in BUFFERs and AGV in different layers.V.DEVELOPMENTANDIMPLEMENTATIONTo settle diversified simulation requirements of an enterprise consisting multi-factories andinstitutes, a model based on eM-Plant platform is realized. Among them one typical machining factory

tasks, product processes and resources in the former instance are organized internally or imported from independent database, while in the latter data have to be imported from related system or by middle database. Like other integration applications, data mapping technologies are employed to transform external data into internal format, such translating work is done in DM. Main data table used in PLS are listed in Table 2. Readers can refer to fig. 4to see their format and instances. Table 2: formatted internal data table Information saved in MODINFO, TRACKINFO and MODLINK is used for generating hierarchical structure of user model in build-time view. Coordinate and status of resources, tasks and processes are dedicated to model running in run-time view. The most earliest production time and the most latest delivery time in tasks can be seemed as the actual start and end running time of simulation respectively. Work pieces in tasks are created in top layer, then they are transferred directly layer by layer along selected routes decided by control rules, or transported by AGV if it is considered in user model; work pieces arrive at machining equipment (EM) and are stored in BUFFERs waiting for machining; The machine selects work piece for machining via selection rules; After machining is finished, the part will be uploaded layer by layer and in the same way downloaded to destination machine responding for its next process. Generally, division of labor in workshop in factory or in cell in workshop is clear, and assignment of task and material transportation route is decided. But when task arrived at cell layer, selection of machining equipment is uncertain according to different schedule rules. Thus, in static model, material transportation is mainly realized in up layers (FM and SM), while middle layer (CM) is used for task assignment, and under layer (EM) only responds for machining. This character can also be adapted to dynamic model. Hence, logic of task assignment of material transportation and machining can be embedded in the joint of layers, that can be easily realized in BUFFERs and AGV in different layers. V. DEVELOPMENT AND IMPLEMENTATION To settle diversified simulation requirements of an enterprise consisting multi-factories and institutes, a model based on eM-Plant platform is realized. Among them one typical machining factory

is illustratedhereasanexample.Fig.4andFig.5areinternal datatableandusermodel separatelyMODLINKTaskListframelismeparertfrareNametastoersyye quntay release datedue dateFristwachiningworkshop国402006/04/1908:000002006/08/19184900.FactpartotPrefoieMachiringworishop0402006/0419.00:00:00.0002006/0/191649:00.FactoyJoart0420060000020/00FSpecisntachiningWorkshop30mpartoe2006/04/1908:000002006/00/1918:49:00.FatoyGeirdnglachiningworkshoo20ot09s1070217002006/04/19 08:00:00002006/08/1918:49:00.Frihiingwokshopn20pet11200/41900006//19:00gwors102900pat12402006/19:0000000/9400/ogwrls17010patt3PreformMachiningworisho:st107068700MasterTablePreformMschiringwiorisho#107071700entbtypeJoperatoneglanopawtosbe11Eat04tie12Nested routetablesetup tineoeratonprocessngtimestationbyepartoo0.0013:20.000016+40.00001st51pert092.0033:20.00004140.0000bt5LNested selectable equipment tablepastn3.001:06:40.0000123:20.0000modt.enframe.Factory.PrefcemMacinngworkshp.st10706670pat424.0029:-20.000036140.0000mods.Userftame.Factory.PreformMachiningWorkshop.st107069700parta315.0024:00.000030:00.0000modt.UserFrame.Factory.PreformMachiningwoekshop.st107071700MODINFOItranetoneMrameCrignIrameobixFactory200200MdMolkmoffFayFnishachiningworishopa1945Modelterplstes.Modoks.C4modi.Userframe.Factory.FirishMadiningworkhopPreformMachiringworkshep140160ModelTenplates.Models.CM.modl.Userframe.Factory.PreformMachiringworkshopSpeclawachiningwarkshop164154ModelTenplstesModels.CMmodtUseifrane.Factory.SpeciamMachiningworkshopGeindnachiningworkshop115270Modeemlstes.Moels.OmodUsefameFactoryGdngMaingWorhpt107021700206387ModelTemplates.Models.D4modl.seeframeFactory.FrishMaciningwelishop.st10702170010702970051945ModeiTenpistes.Models.EM.modtUserFrane.Factory.FinishMachiningWorkshop.stt07029700107048700567146ModenTempistes.Models.EM.modt.Userfrare.Factory.PreformMachiningworkshop.st107048700107068700519249ModeiTenplstes.Modebs.EM.modlUsefrane.Factory.PreformMochiningworkshop.s107068700107071700S84300ModelTerplstesModels.EM.mod1.Usefrane.Foctory.PrefomMachiringworikshop.st107071700Basic modeling objectUsermodeling objectotice:There're three level organization:factory,unit andequipnent in our corporaticnyandthey'reabstracted into SM CMandEMseparatelyin ourmodel.Fig. 4 Instances of internal data tableOutBufferEquipment+FinishaLayout ofmachiningpreform machiningworkshopworkshopDAGVInBufferPreformTransportermachiningworkshopAGVTrackFGrindingSpecialmachiningmachiningworkshopworkshopFig.5The user modelInternal data tables includes task, process and layout table, and user model is automaticallygenerated and running by data-driven method using internal data. There are finish, perform, special